Selecting Steel Beams

Pick the lightest beam that's strong enough, then make sure it isn't too bouncy.

In Steel Beam Flexural Strength, a 30-ft floor beam needed 360 ft-kip, and a W18X50 passed. It works. But the Handbook's Table 3-2 lists three shapes that carry about that much: a W14X61 good for 383 ft-kip, the W18X50 for 379, and a W21X48 for 398.

They weigh 61, 50, and 48 lb/ft. The deepest is the strongest, and also the lightest. Steel is bought by the pound, so on a floor of fifty of these beams, the W21X48 instead of the W14X61 saves 19,500 lb, almost ten tons of steel, for a floor that's no weaker.

Depth buys strength cheaply

Why would the deepest beam be the lightest? At the plastic moment, each flange is fully yielded and carries a force FyAfF_yA_f, one squeezed and one stretched, about d−tfd - t_f apart. That's a couple, so the flanges' share of ZxZ_x is their area times the lever arm, Af(d−tf)A_f(d - t_f), and the web adds the rest.

W14X61 61 lb/ft 383 ft-kip W18X50 50 lb/ft 379 ft-kip W21X48 48 lb/ft 398 ft-kip
Three shapes of about the same strength, drawn to one scale
Below each shape: its weight and its ϕbMpx\phi_b M_{px} from Table 3-2. Dimensions from Table 1-1.

The W14X61's flanges are 6.44 in.² each, on a lever arm of 13.3 in. The W21X48's are just over half that, 3.50 in.², but on a 20.2-in. arm, and its deeper web makes up the difference. Weight is just area: at the Handbook's 0.282 lb/in.³ for steel, each square inch of section weighs 3.4 lb per foot. The W21X48's flanges have 5.9 in.² less steel, about 20 lb/ft, and its taller web adds back only 2.2 in.², about 7 lb/ft: the Hook's 13 lb/ft. Depth is cheap because it costs only web, and the web is thin. It can't grow forever, though, because a floor has only so much room.

Reading Table 3-2

The Handbook's Table 3-2, W Shapes – Selection by ZxZ_x, lists W-shapes of 50 ksi steel from largest ZxZ_x to smallest, with ϕbMpx=0.90(50)Zx/12\phi_b M_{px} = 0.90(50)Z_x/12 already worked out in ft-kip, so you enter it with MuM_u directly. It also lists IxI_x and the design shear strength ϕvVnx\phi_v V_{nx}, for later. Blank rows split it into groups, and each group's first shape is in bold.

Three groups from the middle of Table 3-2
ShapeZxZ_x (in.³)ϕbMpx\phi_b M_{px} (kip-ft)IxI_x (in.⁴)ϕvVnx\phi_v V_{nx} (kips)
W21X48107398959217
W16X57105394758212
W14X61102383640156
W18X50101379800192
W10X7797.6366455169
W12X6596.8356533142
W21X4495.4358843217
W16X5092.0345659185
W18X4690.7340712195
W14X5387.1327541155
W12X5886.4324475132
W10X6885.3320394147
W16X4582.3309586167
W18X4078.4294612169
W14X4878.4294484141
W12X5377.9292425125
W10X6074.6280341129

Adapted from the FE Reference Handbook, AISC Table 3-2: W Shapes – Selection by Zx. Three of its nine groups, without the columns for unbraced beams.

The Handbook never says what bold means, but the table shows it: every other shape in a group is heavier than its bold shape and no stronger, and every shape above a bold one weighs at least as much. So a bold shape is the lightest beam with at least its strength, and selection takes three moves:

  1. Read up the ϕbMpx\phi_b M_{px} column from below to the first shape with ϕbMpx≥Mu\phi_b M_{px} \ge M_u.
  2. Keep reading up to the nearest bold shape, or stop if you're on one. That's the lightest beam strong enough.
  3. Add the beam's own weight and check again.

The Hook's beam carries a factored 3.2 kip/ft over 30 ft, not counting its own weight, so Mu=360M_u = 360 ft-kip. Reading up, the W12X65's 356 falls short and the W10X77's 366 is enough. Stopping there is the trap: the W10X77 works, at 77 lb/ft. Every shape above it in the group works too, and the bold W21X48 is the lightest.

Use ϕbMpx\phi_b M_{px} as printed. It already allows for the few shapes that aren't compact, such as the W21X48, whose 398 is a little under 0.90(50)(107)/12=4010.90(50)(107)/12 = 401. It holds only for 50 ksi, though. For A36, find Zx,req=Mu/ϕbFyZ_{x,\text{req}} = M_u/\phi_bF_y, with MuM_u in in.-kip, and read up the ZxZ_x column instead; the bold shapes still mark the lightest.

The beam's own weight

MuM_u usually arrives without the beam's weight, since nobody knew the beam yet. Once you have one, add 1.2 times its weight to wuw_u: the Hook's 3.2 kip/ft becomes 3.2+1.2(0.048)=3.2583.2 + 1.2(0.048) = 3.258 kip/ft, and Mu=366.5M_u = 366.5 ft-kip, still under 398. Self-weight is a percent or two of the load, so it matters only when MuM_u sits just under a bold shape's strength, and then it pushes you up a group.

Worked Example When the beam's weight tips the choice
1 of 2

A simply supported floor beam of 50 ksi steel spans 30 ft with full lateral support. It carries a factored uniform load of 2.6 kip/ft, not counting its own weight. Select the lightest W-shape from Table 3-2.

When the depth is limited

The number after the W is the nominal depth in inches, a family name close to the true depth dd: Table 1-1 gives 13.9 in. for the W14X61. To save headroom, a question may limit the beam to, say, a nominal 14 in., which allows the W14s and anything shallower.

The bold shortcut is out now, since the bold shape is likely too deep. But within one nominal depth, weight climbs with ZxZ_x, so reading up from the first adequate shape, the first W14 you reach with enough strength is the lightest W14 that works, and likewise for each shallower depth. Take the lightest of those. For the Hook's beam, a 16-in. limit gives the W16X57 and a 14-in. limit the W14X61: 48, then 57, then 61 lb/ft. Headroom costs steel.

Checkpoint

A floor beam of 50 ksi steel with full lateral support must carry Mu=315M_u = 315 ft-kip, including its own weight, and headroom limits it to a nominal depth of 14 in. Which W-shape from Table 3-2 should you select?

A.
W10X68
B.
W14X48
C.
W21X44
D.
W14X53
E.
W16X50

Shear: the web does the work

So the Handbook's shear strength gives the web all of it, at one uniform stress, over a strip as deep as the whole beam: Aw=d twA_w = d\,t_w, "the overall depth times the web thickness." Steel yields in shear at about 0.6Fy0.6F_y, so

Vn=0.6FyAwCv1V_n = 0.6F_yA_wC_{v1}

with Cv1=1.0C_{v1} = 1.0, and the Handbook's Shear section gives ϕv=1.00\phi_v = 1.00 for the webs of rolled I-shaped members, not bending's 0.90:

ϕvVn=1.00 (0.6Fy d tw)≥Vu\phi_v V_n = 1.00\,(0.6F_y\,d\,t_w) \ge V_u
d tw Aw = d tw τ flanges: little shear web: nearly level 0.6Fy elastic Handbook model NA
Shear in a W-shape, and the Handbook's model of it
Not to scale. The web is drawn thicker than true.

Table 3-2 prints this as ϕvVnx\phi_v V_{nx}, for 50 ksi; for another grade, compute it from dd and twt_w. The one exception in the Handbook's tables is the W24X55, whose note v marks a web too slender to count on yielding before it buckles, so its tabulated value already uses ϕv=0.90\phi_v = 0.90. ASD divides by the Handbook's Ω=1.50\Omega = 1.50 for shear, written Ωv\Omega_v here, and compares Vn/ΩvV_n/\Omega_v with the shear from service loads.

Why shear rarely governs

Under a uniform load, Vu=wuL/2V_u = w_uL/2 and Mu=wuL2/8M_u = w_uL^2/8, so Vu/Mu=4/LV_u/M_u = 4/L: the longer the span, the less shear comes with each ft-kip of moment. The Hook's W21X48 sees Vu=3.258(30)/2=48.9V_u = 3.258(30)/2 = 48.9 kips against 217, and shear would govern it only on a span under 4(398)/217=7.34(398)/217 = 7.3 ft. In shear alone, a span can carry wu=2ϕvVn/Lw_u = 2\phi_vV_n/L.

Worked Example A floor beam in shear
1 of 3

A W18X50 of 50 ksi steel spans 20 ft, simply supported, under a factored uniform load of 6.0 kip/ft, including its own weight. Find its design shear strength from dd and twt_w, and check it in shear. Then find the largest factored uniform load the span could carry in shear alone.

A heavy load near a support

A point load breaks the 4/L4/L rule. With PP a distance aa from the left support and bb from the right, the near reaction is R1=Pb/LR_1 = Pb/L, and the moment under the load is R1a=Pab/LR_1a = Pab/L. Slide the load toward the support, and R1R_1 climbs toward all of PP, while the moment shrinks with its lever arm aa.

P R1 = Pb/L R2 = Pa/L a b L V Pb/L −Pa/L M Pab/L
A point load near one support, with its shear and moment

So a short beam that picks up a column near one end needs VuV_u at the near support checked against ϕvVnx\phi_v V_{nx}, even when bending passes easily.

Checkpoint

A W12X40 of 50 ksi steel spans 6 ft, simply supported, and carries a factored point load of 150 kips 1.5 ft from its left support. Its own weight is negligible here. Table 3-2 gives ϕbMpx=214\phi_b M_{px} = 214 ft-kip and ϕvVnx=106\phi_v V_{nx} = 106 kips. Which is right?

A.
It passes both checks: Vu=75V_u = 75 kips and Mu=168.8M_u = 168.8 ft-kip.
B.
It fails in shear: Vu=112.5V_u = 112.5 kips exceeds 106 kips, though Mu=168.8M_u = 168.8 ft-kip passes.
C.
It fails in flexure: Mu=225M_u = 225 ft-kip exceeds 214 ft-kip.
D.
It fails both checks: Vu=112.5V_u = 112.5 kips exceeds 106 kips, and Mu=225M_u = 225 ft-kip exceeds 214 ft-kip.

Too bouncy: deflection

A beam can be plenty strong and still sag enough to crack the plaster under it or bounce underfoot. That's a serviceability limit state, from Loads and Load Combinations: nothing comes down, but the floor doesn't do its job on an ordinary day. So the check uses service loads, with no load factors.

Often it uses the live load alone. The dead load's sag happens once, before the partitions and ceilings go in. What cracks them is the extra sag each time the room fills up. The Handbook prints no deflection limits, so the question states one, most often L/360L/360 under live load: 1.00 in. on a 30-ft span. Set the sag against the limit:

5wL4384EI≤L360⟹Ireq=5wL4384E (L/360)\frac{5wL^4}{384EI} \le \frac{L}{360} \quad\Longrightarrow\quad I_{\text{req}} = \frac{5wL^4}{384E\,(L/360)}
Worked Example The Hook's beam under its live load
1 of 2

The Hook's W21X48 spans 30 ft and carries service loads of 1.0 kip/ft dead, plus its own weight, and 1.25 kip/ft live. Its live-load deflection is limited to L/360L/360. Find the minimum moment of inertia, and check the W21X48.

Turning the check around

The same inequality gives the longest span or the largest load. The limit L/360L/360 grows with the span, so one LL cancels, and the span enters as L3L^3:

Lmax=384EI5(360) w3wmax=384EI5(360) L3L_{\text{max}} = \sqrt[3]{\frac{384EI}{5(360)\,w}} \qquad w_{\text{max}} = \frac{384EI}{5(360)\,L^3}

A span 10% longer sags 46% more but is allowed only 10% more, which is why deflection tends to take over on long, lightly loaded beams.

Checkpoint
1 of 3

A W18X40, with Ix=612 in.4I_x = 612\ \text{in.}^4 from Table 1-1, spans 32 ft, simply supported, under a service live load of 0.9 kip/ft. Does it meet a live-load limit of L/360L/360?

A.
Yes: its sag of 0.24 in. is within 1.07 in.
B.
No: its sag of 1.91 in. exceeds 1.07 in.
C.
Yes: its sag of 1.20 in. is within L/240 = 1.60 in.
D.
No: its sag of 1.20 in. exceeds 1.07 in.

Stiffness, not strength

A stronger steel won't fix a beam that sags too much, because FyF_y appears nowhere in 5wL4/(384EI)5wL^4/(384EI). EE does, and every structural steel has the same one: the Handbook's 29,000 ksi for steel design, and the same 29.0×10329.0 \times 10^3 ksi for A36 in its Table 2. A stronger steel follows the same stress-strain line and only goes higher before it yields.

0 10 20 30 40 50 60 0 0.001 0.002 0.003 strain, ε stress, σ (ksi) 50 ksi steel A36 (36 ksi) slope E = 29,000 ksi for both
Two grades of steel, one modulus

Under service loads a beam stays on that shared line, so the same shape in either grade sags exactly the same. The cure is a larger II, and since the flanges supply most of it at about d/2d/2 from the axis, II grows with roughly the square of the depth. Go deeper.

Checkpoint
1 of 2

A floor beam carries service loads of 0.9 kip/ft dead and 1.2 kip/ft live. Its deflection is limited to L/360L/360 under live load. What uniform load goes into the deflection check?

A.
2.1 kip/ft, from D + L
B.
3.0 kip/ft, from 1.2D + 1.6L
C.
1.92 kip/ft, from 1.6L
D.
1.2 kip/ft, the live load alone

Recap

  • Depth buys strength cheaply, because the flanges' lever arm is the depth, so the lightest adequate W-shape is usually a deep one.
  • In Table 3-2 (50 ksi only), read up to the first shape with ϕbMpx≥Mu\phi_b M_{px} \ge M_u, then up to the nearest bold shape, and recheck with 1.2 times its weight. Under a depth limit, take the lightest of the first shapes of each allowed depth.
  • ϕvVn=1.00(0.6Fy d tw)\phi_v V_n = 1.00(0.6F_y\,d\,t_w), and ASD uses Vn/1.50V_n/1.50. Shear governs only on very short spans or under a heavy load near a support.
  • Deflection uses service loads, often the live load alone, against a stated limit such as L/360L/360, in kips and inches.
  • Every steel has E=29,000E = 29{,}000 ksi, so only a larger II, not a higher grade, reduces the sag.